SOCIAL LEARNING THEORY: UNDERAGE DRINKING, BLACK ... [PDF]

SOCIAL LEARNING THEORY: UNDERAGE DRINKING, BLACK MARKET ASSOCIATIONS, SUBSTANCE USE AND DEVIANCE

By HEATHER STEWART

A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2010 1

© 2010 Heather Stewart

2

To my mother and all those who nurtured my intellectual curiosity, academic interests, and sense of scholarship throughout my lifetime, making this milestone possible

3

ACKNOWLEDGMENTS I wish to thank my mentor, chair and intellectual inspiration, Ronald L. Akers, his guidance and support has been invaluable. I wish to thank my supervisory members (Lonn Lanza-Kaduce, and Marv Krohn), their expertise truly made this possible. Finally, I wish to thank my fellow criminology graduate students for the mutual support during the frequent stressful times.

4

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................................... 4 ABSTRACT .......................................................................................................................................... 8 CHAPTER 1

PURPOSE OF THE STUDY ..................................................................................................... 10 Research Questions ..................................................................................................................... 10 Purpose ......................................................................................................................................... 10 Hypotheses................................................................................................................................... 11

2

INTRODUCTION....................................................................................................................... 13

3

LITERATURE REVIEW ........................................................................................................... 16 Theoretical Rationale and Literature Review: A Social Learning Perspective on “Black Market Associations” .............................................................................................................. 16 Prior Research: The Legal Drinking Age .................................................................................. 18 Prior Research: Underage Drinking ........................................................................................... 20 Prior Research: Social Learning Theory and Underage Drinking ........................................... 23

4

STUDY PROCEDURES ............................................................................................................ 25 The Present Study........................................................................................................................ 25 Methodology................................................................................................................................ 27 Participants .................................................................................................................................. 28 Materials and Procedures ............................................................................................................ 31 Measures of Control Variables ................................................................................................... 33 Measure of Alcohol Black Market Sources ............................................................................... 34 Measures of Dependent Variables ............................................................................................. 34 Marijuana Use ...................................................................................................................... 35 Marijuana Black Market Sources ....................................................................................... 35 General Deviance................................................................................................................. 36 Specific Black Market Deviance ........................................................................................ 36 Measures of Social Learning Variables ..................................................................................... 39 Differential Reinforcement ................................................................................................. 39 Definitions/ Attitudes .......................................................................................................... 40 Differential Association ...................................................................................................... 42 Imitation/ Modeling ............................................................................................................. 45 Data Analysis............................................................................................................................... 47

5

5

RESULTS .................................................................................................................................... 54 Alcohol Black Market Sources................................................................................................... 54 Marijuana Black Market Sources ............................................................................................... 55 Marijuana Use ............................................................................................................................. 58 General Deviance ........................................................................................................................ 60 Support for the Hypotheses ........................................................................................................ 61

6

CONCLUSION ........................................................................................................................... 63 Discussion .................................................................................................................................... 63 Limitations and Implications ...................................................................................................... 64

APPENDIX A

SOCIAL LEARNING MODEL ................................................................................................. 66

B

IRB PROTOCOL FORM ........................................................................................................... 68

C

INFORMED CONSENT FORM ............................................................................................... 73

D

DEBRIEFING FORM ................................................................................................................ 76

E

INSTRUMENT ........................................................................................................................... 78

F

CORRELATION MATRICES................................................................................................. 113

G

BIOGRAPHICAL SKETCH .................................................................................................... 124

6

LIST OF TABLES Table page 4-1 Participant Characteristics ..................................................................................................... 32 4-2

Descriptive Statistics .............................................................................................................. 47

5-1

Model 1 ................................................................................................................................... 54

5-2

Model 2 and Model 3 ............................................................................................................. 56

5-3

Model 4 and Model 5 ............................................................................................................. 58

5-4

Model 6 and Model 7 ............................................................................................................. 60

7

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts in Criminology, Law & Society SOCIAL LEARNING THEORY: UNDERAGE DRINKING, BLACK MARKET ASSOCIATIONS, SUBSTANCE USE AND DEVIANCE By Heather Stewart August 2010 Chair: Ronald Akers Major: Criminology, Law and Society This research is the first part of a multiphase study which focuses on underage drinking, methods of obtaining alcohol while underage, and related deviance. Specifically this study examines the applicability of Akers’ social learning theory in this context. Youth under the legal drinking age are illegally obtaining alcohol. Previous research has revealed that underage drinkers sometimes utilize persons they are not closely associated with, or strangers of legal drinking age, to procure alcoholic beverages (Lanza-Kaduce and Richards, 1989). In the literature this method of procuring alcohol has been labeled “black market.” In this research, I examine black market associations made while obtaining alcohol illegally in relation to other black market associations, other black market deviance, and deviance in general. This research utilizes social learning theory as a framework within which these black market peer associations are examined, specifically whether or not they facilitate further deviance, and more explicitly the extent to which they increase the likelihood of other black market deviance. The study sample was 408 undergraduates drawn from the departmental participant pool at a major southeastern university. Data were collected through an online survey instrument and analyzed with multivariate statistical techniques. Results indicated that having black market peer associations

8

for alcohol was predictive of further black market associations, substance use in general, and general deviance. As hypothesized, these relationships were diminished when entered into the equation with the social learning variables. Black market peer associations are viewed as a specific type of differential peer association, so the variance is explained by the model including the social learning variables.

9

CHAPTER 1 PURPOSE OF THE STUDY Research Questions Research questions were as follows: Can Alcohol Black Market Sources be predicted using social learning variables?; Does the use of Alcohol Black Market Sources have a significant effect on the use of Marijuana Black Market Sources, and is that relationship diminished when social learning variables are present?; Does the use of Alcohol Black Market Sources have a significant effect on the use of marijuana, and is that relationship diminished when social learning variables are utilized?; Does the use of Alcohol Black Market Sources have a significant effect on General Deviance, and is that relationship diminished when social learning variables are present? For the purposes of this study, using black market sources was defined as the utilization of any person personally unknown to the participant trying to obtain the illegal substance (be it alcohol, drugs or otherwise). Sources were only labeled black market if the participant indicated that the source was a stranger or another student that they did not really know. The illegal substances in this study included alcohol for underage drinkers and marijuana. Purpose This research attempted to examine the behavior of underage drinkers, specifically the manner in which they procured alcohol, and if that included black market associations, whether those participants were more likely to be involved in deviant behavior, including other black market deviance. Researchers intended to examine the black market relationships developed by underage drinkers in the context of Akers’ social learning theory in an attempt to indicate whether or not those relationships were simply a specific type of differential association, or if it was something entirely different. To date, there has been no research examining black market sources utilized by underage drinkers and the possible link to further deviance and the

10

development of other black market sources. There was one published study that looked at the methods underage drinkers use to obtain alcohol (Lanza-Kaduce and Richards, 1989), but aside from determining those methods, the literature is lacking. Thus the current study is unique. Black market sources developed to obtain alcohol by underage drinkers has not yet been looked at. Furthermore, social learning theory has never been applied in this specific manner. The internet survey was designed specifically to measure both black market sources and all four social learning variables, utilizing completely different survey items for black market sources and differential association. These two measures are never used tautologically. It is important to understand the link between underage drinking, substance use and deviance. If there is a link between Alcohol Black Market Sources, Other Black Market Sources and General Deviance, it could provide some insight for future intervention actions and policy changes. Further, conducting a study that can potentially expand the empirical support for social learning theory, and provide a unique understanding of black market associations, may allow for a full test of the social structure social learning (SSSL) model in the future. This second study would test these same variables on both a national level and an international level, with a survey specifically designed to measure all eight variables in the SSSL model, possibly for the first time. Hypotheses Based on the findings from extant research and the theoretical framework of social learning theory, our study addressed four hypotheses: 1) The use of Alcohol Black Market Sources is predicted by social learning variables; 2) The use of Alcohol Black Market Sources has a significant effect on use of Marijuana Black Market Sources, but that effect will be reduced and become non-significant when measures of social learning variables are entered in a multiple regression model; 3) The use of Alcohol Black Market Sources has a significant relationship 11

with use of marijuana (Marijuana Use), but that relationship will become non-significant when measures of social learning variables are entered in a multiple regression model; 4) The use of Alcohol Black Market Sources has a significant relationship with other measures of deviant behavior (General Deviance), but that relationship will become non-significant when measures of the social learning variables are entered in a multiple regression model.

12

CHAPTER 2 INTRODUCTION Much research has been conducted in the area of youth alcohol use. This research covers everything from deviance associated with early onset drinking to the negative results it can have on health. College students comprise a large portion of the sample populations from which this area of research draws (Wechsler, Davenport, Dowdall, Moeykens & Castillo, 1994, Wechsler, Dowdall, Maenner, Gledhill-Hoyt, & Lee, 1998; Wechsler, Nelson & Kuo, 2002). Underage drinking, binge drinking, and drinking in general, are thought by mainstream America to run rampant on college campuses, and to some degree they are correct. Drinking is, in fact, a problem faced by many college campuses (Akers & Jenson, 2003; Boeringer, Shehan, & Akers, 1991; Durkin, Wolfe & Clark, 2005; Durkin et al., 1996; Hood, 1996; Lanza-Kaduce, Capece & Alden, 2006; Lanza-Kaduce & Capece, 2003; Mayer, Foster, Murray & Wagenaar, 1996; Sun & Longazel, 2008; Wechsler et al., 2002). Research indicates that college students are significantly more likely to partake in heavy drinking than their peers not enrolled in college, who are the same age (Bachman, O’Malley & Johnson, 1984; O’Malley & Johnson, 2002; Sun & Longazel, 2008). More importantly, both and on and off college campuses, drinking has been linked to a plethora of other problems. Rape is one of the most serious issues related to drinking on college campuses, especially acquaintance rape (Durkin et al., 2001; Boeringer, Shehan, Akers, 1991; Abbey, 1991; Koss & Dinero, 1989; Muehlenhard & Linton, 1987). Drunk driving, the very catalyst for raising the LDA (legal drinking age), is also observed as a result of drinking on college campuses (Durkin et al., 2001; Robinson, Roth, Gloria, Keim & Sattler, 1993; Werch, Gorman, & Marty, 1987; Engs & Hanson, 1988; Saltz & Elandt, 1986). Students with drinking

13

problems have a propensity towards low self control and indicate increased deviant behavior (Gibson, Schreck & Miller, 2004; Haberman, 2001; Piquero, Gibson, & Tibbetts, 2002). Overall aggression, including fighting, bullying, violence and property damage, is another concern that is positively correlated with drinking on college campuses (Durkin et al., 2001; O’Hare, 1990; Engs & Hanson, 1988; Werch, Gorman & Marty, 1987; Wechsler & Rohman, 1981). Poor academic performance is noted as a result of drinking by college students (Lanza-Kaduce, Capece, & Alden, 2006; Durkin et al., 2001; Rapaport, 1993; Haberman, 2001; Johnston, Bachman, & O’Malley, 1992; Saltz & Eldant, 1986; Wechsler & Nelson, 2008). And finally, researchers combining statistics from three national surveys, and a myriad of other data sources, estimated that approximately 1,700 college students die every year from alcohol related accidents/ injuries (Higson, Heeren, Zakocs, Kopstein, & Wechsler, 2001). Thus, drawing samples from college campuses makes sense, as drinking is a problem, and there are both underage and legal age students who drink in these populations. Substance use other than alcohol is also a prevalent problem on college campuses across the United States (Gledhill-Hoyt, Lee, Strote & Wechsler, 2000; Johnston, O’Malley, Bachman & Schulenberg, 2004; McCabe, Schulenberg, Johnston, O’Malley, Bachman & Kloska, 2005; Haberman, 2001; Strote, Lee & Wechsler, 2001). Marijuana is the most prevalent illicit substance utilized by college students (Johnson et al., 2004). It is followed by hallucinogens, amphetamines, cocaine and ecstasy or MDMA, though the use of MDMA has been increasing while other illicit drug use has remained fairly stable (Johnson et al., 2004; Larimer, Kilmer & Lee, 2005; Strote, Lee & Wechsler, 2002). While some studies cite health reasons, increased risk taking behavior, and poor

14

academic performance as reasons for studying illicit substance use (Larimer et al., 2005; Strote et al., 2002), others allude to the fact that college students are at an age where they are without parental control, usually for the first time, able to afford illicit drugs, and have a “tendency… to try new, previously prohibited behaviors” (Gledhill-Hoyt et al., 2000: 1656).

15

CHAPTER 3 LITERATURE REVIEW Theoretical Rationale and Literature Review: A Social Learning Perspective on “Black Market Associations” While this literature has highlighted the links between underage drinking and other problems, little of it has focused on the unintended problematic consequences of underage drinkers procuring alcohol through illicit means, especially that which involves contact with strangers and others, which Lanza-Kaduce and Richards(1989) refer to as “black market associations.” The present research investigated the possible devianceenhancing effects of engaging in such “black market” behavior such as using the same techniques to procure other substances that are illegal for all ages such as marijuana, as well as other forms of deviant and illegal behavior. The argument here is that using such black market sources for underage acquisition of alcohol and its links to black market sources for other substances and other deviant behavior can be understood by referencing social learning theory concepts and variables -- differential association, differential reinforcement, imitation, and definitions social learning theory (Akers, 1998). Although there is research on the connections between illicit procurement of drugs and violent and other criminal behavior, to date there is no research on the connections between the black market procurement of alcohol by minors and these other forms of crime and deviance. Social learning theory has been developed by Akers and some of his colleagues. It has been well supported in studies dealing with substance use/ abuse, crime, and deviance. Indeed, social learning theory is one of the most empirically supported theories in criminology (Akers, 1973, 1977, 1985, 1994, 1998; Akers, Krohn, Lanza-Kaduce, and Radosevich, 1979; Burgess and Akers, 1966; Akers et al., 1989; Akers and Sellers 2008; 16

Akers and Jensen, 2003; 2006). However, the present research is the first to apply social learning theory to use of “black market” sources by minors procuring alcohol. In social learning theory, behavior is viewed as elicited by the physical and social environments in which the person is located at any given time and through time (Akers, 1998; Akers & Sellers, 2008; Bandura, 1973, 1977; Braukmann, Fixen, Phillips & Wolf, 1975). Social learning theory, as defined by Akers, encompasses four main variables that explain how behavior, conforming or deviant, is learned and maintained: differential associations, definitions, differential reinforcement and imitation. Differential association refers to the direct and indirect association with others who engage in and support different types of behavior. These associations “provide the major social contexts in which all the mechanisms of social learning operate” (Akers and Sellers, 2008). Differential peer associations are one of the strongest predictors of delinquency in youth (Agnew, 1991; Akers, 1998; Brank, Lane, Turner, Fain & Sehgal, 2008; Dishion, McCord & Poulin, 1999; Longshore, Chang & Messina, 2005). Definitions, Akers contends, are the meanings, beliefs, and attitudes one has toward a particular behavior. The more positive the definition of the behavior, the more likely the behavior is to occur – be it conforming or deviant. Differential reinforcement is the balance of perceived rewards and punishments consequential to a behavior. The more rewarding one views the behavior the more likely that person is to partake in it and repeat it in the future, thus it works both to create or dissuade, and maintain or desist behavior. There are two types of reinforcement, negative and positive. Negative reinforcement strengthens a behavior by rewarding a behavior that removes something negative, such as putting on sun glasses to alleviate the glare of the sun, or using a drug to alleviate a withdrawal symptom. Positive

17

reinforcement strengthens a behavior by presenting something positive, much like a pat on the back after hitting a home run, or verbal approval by a friend after certain a behavior has occurred. Punishment works in a similar fashion. Negative punishment is the removal of something pleasant after a behavior occurs that weakens that behavior (being grounded for a poor report card), while positive punishment is the presentation of an adverse stimulus subsequent to a behavior, also weakening the target behavior (a hangover after an evening of drinking). Finally, imitation is the fourth learning variable. It is more likely to affect the acquisition of novel behavior, but continues to have some effect in the maintenance of behavior. It is exactly how it sounds; one imitates or models the behavior after observing another act out. These four variables are micro level variables, meaning they apply specifically to the individual or a small group of individuals (Akers, 1973, 1977, 1985, 1994, 1998; Akers, Krohn, Lanza-Kaduce, and Radosevich, 1979; Burgess and Akers, 1966; Akers et al., 1989; Akers and Sellers 2008). Prior Research: The Legal Drinking Age Many studies have looked at the LDA and its implications. Wagenaar et al. completed a meta-analysis of all the literature that focused on the LDA from 1960 to 2000 (2002). The overall tone of this analysis confirms what both Lanza-Kaduce and Richards found in 1989 from U.S. statistics, and what Nickerson found from Canadian data in 2001. Both lowering and raising the legal drinking age have had unintended consequences. In fact, the goal of raising the legal drinking age was to ameliorate problems associated with emerging adult/ adolescent alcohol consumption. Lowering the age is associated with increased consumption among younger people and raising the age has unintentionally played a role in other problems (Wagenaar et al., 2002; LanzaKaduce and Richards, 1989; Nickerson, 2001). 18

Lanza-Kaduce and Richards experienced a unique opportunity for research when the legal drinking age was changed from 19 to 21 in Florida. All of the 19 year olds born after July 1, 1966 would have to wait until their 21st birthday to drink, while anyone born before that date was grandfathered in. Thus, researchers had access to 19 year olds that the new law affected, and 19 year olds that it did not affect (1989). Lanza-Kaduce and Richards had several research questions in mind, and among them was: whether or not raising the LDA led to derivative law breaking, and what impact it had on the social context of drinking. They had some interesting findings. Underage drinkers were much more likely to utilize persons they are not closely associated with, or strangers (black market associations), to obtain alcohol, and to do so “they must associate with others who are willing to break the law” (1989). Underage drinkers were also shown to have used false ID’s to obtain alcohol. Both of these methods of procuring alcohol are considered derivative law breaking. Researchers also found that the social context of drinking changes. Drinkers of legal age are more likely to drink in public venues, while underage drinkers preferred their dwellings. And finally, the underage drinkers were more likely to drink with unfamiliar persons (Lanza-Kaduce & Richards, 1989). This line of research is expanded upon in the present study. Other research has found that when the drinking age was lowered from 21 to 18 the levels of drinking and problems associated with drinking increased among those just below that age (16-17); when the age was increased to 21 “the age at which problems were most visible rose again to 19-20, just below the drinking age” (Akers, 1992:220).

19

Prior Research: Underage Drinking There is a plethora of research concerning adolescent and young adult substance use and alcohol consumption. Some of the ideas for underage drinking research are drawn from studies completed on adult drinking patterns. Situational factors have been the focus of several studies, and have been shown to be related to alcohol use. (Miller et al, 2005; Mayer et al., 1996; Casswell, Zhang & Wyllie, 1993; Connolly & Silva, 1992, Donnermeyer and Park, 1995; Gibbons et al, 1986) For both adults and underage drinkers, the setting played an important role in amount of alcohol consumed. Adults tended to drink heavier in public settings such as clubs or bars (Mayer et al. 1996; Harford, 1975; Caetano and Herd, 1988; Harford 1983), while underage drinkers were more prone to heavy drinking in a private setting, such as a residential party or friend’s house (Lanza-Kaduce and Richards, 1989). Also, adolescents who participated in binge drinking were more likely to drink with peers and strangers, while those who did not binge drink reported drinking with their parents or siblings. This finding was most prevalent amongst the females (Mayer et al., 1996). Another study found that club settings, especially those hosting electronic music dance events, were prime public targets for facilitation of youth drug and alcohol use (Miller et al., 2005). Using anonymous surveys, breath tests, and saliva samples, Miller et al. were able to determine that while most underage drinkers showed up with alcohol in their system; some were also able to obtain it from within the club setting (2005). The most prevalent substance used at the electronic music dance events was alcohol, followed closely by marijuana. Stimulants (i.e. ecstasy and cocaine) were third on the list upon entering the club, but moved up to second upon exit (this may be partially due to the method of testing, marijuana only stays in the saliva for fifteen minutes, thus is harder to 20

accurately measure after time has elapsed) (Miller et al., 2005). According to Miller et al., risky behavior is associated with alcohol and drug use in clubs, including overdose, illegal possession of drugs, increases in aggressive behavior, increased risk for victimization, and driving under the influence (2005; Collins, 2004). Access to alcohol by minors has also been the focus of several studies. Alcohol is viewed by a substantial number of minors as easy to access. (McFadden and Wechsler, 1979; Goldsmith, 1988; Klepp, Schmid & Murray, 1996; Smart, Adlaf & Walsh, 1996; Wagenaar et al., 1996; Jones-Webb, Toomey, Wagenaar, Wolfson & Poon, 1997; Mayer et al., 1998; Wagenaar & Toomey, 2002; and Wechsler et al., 2002). Wagenaar and Toomey reported that over 75% of teens that were surveyed reported that alcohol was easy to obtain (2002). Some common methods of acquiring alcohol included: False identification (either forged or an ID belonging to someone of age), attempts to purchase alcohol without an ID, having peers or family members provide it, and having an adult of legal drinking age purchase alcohol on their behalf (see appendix A) (Lanza-Kaduce and Richards, 1989; Wagenaar et al., 1996; Wagenaar and Toomey, 2002; Wechsler et al., 2002). Using false identification to procure alcohol is one of the methods mentioned above. There have been numerous studies that have examined this avenue. Goldsmith explained the methods in which youth can obtain a false ID, which included: falsifying one’s own ID, borrowing or stealing an ID from someone of age, using false documents to obtain an ID, or forging a fake ID (1988). In a study completed by Schwartz, Farrow, Banks & Giesel, 14% of the youth they sampled reported using false identification (1998). This finding confirmed prior research that false identification is in fact a

21

common method employed by youth to purchase alcohol (Goldsmith, 1988; Durkin et al., 1996; Klepp et al., 1996) Schwartz et al.’s study led nicely into a study by Durkin et al. which specifically examined the use of false identification in procurement of alcohol by underage college students (2001). Of the college students they sampled, almost half reported using a fake ID to gain access to alcohol. Greek membership was strongly related to the use of fake ID’s, however, other students used them as well (Durkin et al., 2001). This finding suggests that use of a false ID may be more prevalent among college students, than youth not enrolled in college. Durkin also tested some of the social learning variables (differential associations and definitions), and found that they explained a significant amount of variance, 30% (2001). Though using false identification to obtain alcohol is prevalent in underage youth, the most popular method employed was identified as obtaining alcohol from a person over the legal drinking age, which includes strangers. (Wagenaar & Toomey, 2002; Durkin et al., 2001; Wagenaar et al., 1996; Smart et al., 1996; Klepp et al., 1996; LanzaKaduce and Richards, 1989; Goldsmith 1988; McFadden & Wechsler, 1979). There are several interactions noted in the research in this area. Some underage drinkers use family or friends to obtain alcohol (McFadden & Wechsler, 1979; Lanza-Kaduce & Richards, 1989; Wagenaar et al., 1993; Wagenaar et al., 1996; Klepp et. al., 1996; Smart et al., 1996; Jones-Webb et al., 1997). Many studies have failed to categorize the adults that underage drinkers utilized to procure their alcohol. Lanza-Kaduce and Richards however, found that underage drinkers do utilize strangers, family members and friends over the age of 21 to obtain alcoholic beverages, but the degree to which each category is used is undetermined by research (1989). There is currently no research that thoroughly

22

differentiates the different methods of obtaining alcohol while underage, and more specifically differentiates what relation the sources have, if any, to the underage drinker, when they are utilizing persons over the age of 21 to procure their alcohol. Prior Research: Social Learning Theory and Underage Drinking In addition to Durkin et al.’s study in 1996 mentioned above, there are a number of studies that include measures of social learning in underage drinking and substance use. Durkin et al. completed another study in 2005 evaluating social learning theory in application with binge drinking in college students. They found that, “differential peer associations are by far the best predictor of this behavior [binge drinking].” (2005). Durkin et al. (2005) also found significant effects of definitions, as well as perceived reinforcements on drinking behavior. Hood found that the most influential factor in underage drinking is peer influence, and the “best predictors of experiencing trouble with the police are having a friend who has been arrested…” (1996). Akers, Krohn, LanzaKaduce and Radosevich were able to explain 55% of the variance in the use of alcohol by adolescents, using the social learning variables (1979). Lanza-Kaduce and Capece also completed a study on binge drinking that employed components of social learning theory. Overall, they were testing the social structure social learning model (SSSL), but it encompasses the four learning variables (2003). Akers’ used social learning theory as the foundation for the SSSL model. In addition to the four micro level variables tested by learning theory, he added four structural variables, and hypothesized that the structural variables would be mediated by the learning variables (Akers, 1994; 1997; 1998; 2000; 2008). In Lanza-Kaduce and Capece’s article, researchers found support for Akers’ model, though they did not test it in full. The effects of the structural variables were reduced when the learning variables were controlled for. The social learning variables 23

tested included definitions and perceived reinforcement. Even though only half of the theory was tested, it was still able to explain 16 to 34% of the variance in binge drinking (2003). The SSSL model is tested in part again in 2006, in its applicability to gender, alcohol use and sex (Lanza-Kaduce, Capece & Alden, 2006). In a response to criticism of the SSSL model, researchers compared feminist theory with SSSL in drinking behavior and sex. It was found that the masculinist pattern in drinking before sex was partially mediated by social learning variables (Lanza-Kaduce, Capece & Alden, 2006).

24

CHAPTER 4 STUDY PROCEDURES The Present Study The term “black market associations” as used by Lanza-Kaduce and Richards (1989) may itself be viewed as a specific type of deviant association. Alcohol black market associations are expected to lead to further deviance, both in general and in the form of substance use. Thus, according to the implications of social learning theory, these deviant associations should increase the risk of engaging in further deviance. Social learning theory would view underage drinking, using various legal and illegal substances, obtaining supplies through illicit or black market means, and engaging in other forms of deviance as inter-related and all explicable in terms of the variables implicated in the social learning process. “Black market associations” (Lanza-Kaduce & Richards, 1989) because they involve association with non-conforming others may be correctly viewed as a specific type of deviant association. As such, they may be seen as indirect indicators of the process of differential association in social learning theory. But the meaning and measure of black market associations are simply reports of contact or interaction by the person with others for the specific purpose of using them to help obtain, or as a source for, the substance. They are not the common measure of differential association which is the proportion of deviant and conforming others with whom one is in association. Therefore, they will not be used as measures of differential association in this study. To emphasize that, and to avoid possible confusion with the measures of differential association that are used in this study, “black market sources” may be a better term, and will be the term regularly used here to refer to interaction with others for procurement of alcohol and marijuana. Both are behavior assumed to result

25

from social learning processes of differential association, differential reinforcement, imitation, and definitions. Therefore, social learning variables should be able to account for use of black market sources for both alcohol and marijuana and for the relationship between them. Since those who use marijuana and other drugs are very likely to have previously used alcohol, the expectation is that use of black market sources for alcohol precedes and provides a social context which facilitates, or leads to, a higher probability of using black market sources to obtain marijuana, and should also be related to involvement in other forms of substance use and deviance. Therefore, use of black market sources for alcohol is hypothesized to be predictive of Marijuana Black Market Sources, use of marijuana, and deviance. However, since all of these are hypothesized to result from the same underlying social learning process, it is expected that when measures of the social learning variables of differential association, imitation, differential reinforcement, and definitions are entered into the model, the net effects of the use of black market sources for alcohol on these other forms of deviance will be lower than their main effects and be statistically non-significant. It may be said that this renders the relationships between the Alcohol Black Market Sources and the other forms of deviance spurious, or simply that they are all indicators of the social learning process in substance use, obtaining substances for that use, and therefore correlated with other forms of deviant behavior. As mentioned on pages 11-12, based on this social learning perspective and prior research this study addresses the following hypotheses (they are mentioned here again with reference to the models and tables presented later, relevant to each hypothesis):

26

Hypothesis 1: Use of Alcohol Black Market Sources is predicted by social learning variables (for the findings on this hypothesis see Model 1, Table 5-1, page 54). Hereafter, this variable (Alcohol Black Market Sources) is hypothesized as an independent variable, along with the social learning variables which remain as independent variables in all of the analyses. Hypothesis 2: Use of Alcohol Black Market Sources has a significant effect on use of Marijuana Black Market Sources, but that effect will be reduced and become nonsignificant when measures of social learning variables are entered in a multiple regression model (see Models 2 & 3, Table 5-2, page 56). Hypothesis 3: Use of Alcohol Black Market Sources has a significant relationship with use of marijuana (Marijuana Use), but that relationship will become non-significant when measures of social learning variables are entered in a multiple regression model (see Models 4 & 5, Table 5-3, page 58). Hypothesis 4: Use of Alcohol Black Market Sources has a significant relationship with other measures of deviant behavior (General Deviance), but that relationship will become non-significant when measures of the social learning variables are entered in a multiple regression model (see Models 6 & 7, Table 5-4, page 60). Methodology Thesis Research Protocol Approved by the University of Florida Internal Review Board. The protocol for this research was submitted the UFIRB 02 on September 25th, 2008. The proposal was passed by UFIRB 02 on October 3, 2008 and was issued a protocol number, #2008-U-886. The letter of approval was typed and sent out October 6, 2008. The UFIRB 02 protocol form can be found in Appendix D,

27

followed by the Informed Consent form, and the Debriefing forms in Appendixes E and F. Participants The outcome analysis reported here is based on a cross-sectional survey completed by students in attendance of classes offered through the University of Florida’s Department of Criminology, Law and Society during the fall 2008 and spring 2009 semesters. Students enrolled in criminology classes are given the option to participate in research for a certain number of points in their classes. Participation in research is completely voluntary, and alternative assignments of an equal time investment are offered in all classes requiring participation points. Participants signed up through the Participant Pool, which is an online listing of current research within the department. Students could choose which studies to participate in, as several studies were posted at any one time. An alternative assignment was available for those students who did not wish to participate in research. The survey was posted on a survey website, www.surveymonkey.com which students could get to by signing up for this study through the Participant Pool. SurveyMonkey has two types of subscriptions, free and paid. I chose to use the paid service due to the length of the survey, the number of participants anticipated, and for data encryption to protect the privacy of the participants and their responses. At the very beginning of the survey, the students were provided with an informed consent form to read over, they were required to click “next” to indicate their agreement to participate in the study (see Appendix D). Participants were able to print or save a copy of the informed consent. The Informed consent form listed contact information for researchers, in case respondents had any questions in the future. Upon giving their informed consent 28

by clicking “next,” the participants were directed to the online survey (see Appendix G) and requested to fill it out to the best of their ability. Students were reminded in the informed consent that they are not to put their name on the survey, and there is no blank field for them to do so. The survey was completely anonymous. Upon completing the survey, students were debriefed as to the purpose of the study and provided a description of what the researchers are looking for (see Appendix F), again the participants could print or save a copy of the debriefing form. The Participant Pool is used not only to collect data, but also to introduce the students to research in academia. This questionnaire was designed to ask questions about general demographics, campus environment, underage drinking, methods of obtaining alcohol, substance use, methods of obtaining substances, deviance of friends, friends’ substance use, substance use of other students and other forms of deviance. Several questions from the survey will be used as measures of the social learning variables, while others will be used to measure the outcome variables of interest (substance use, deviance, other black market associations). This survey was developed under the supervision of Ronald Akers, Ph.D. and Lonn Lanza-Kaduce, Ph.D., to ensure proper measurement of the social learning variables and deviance. Most of the attitudinal questions, activity questions, demographic questions, and substance use questions were derived from the questions asked in the CORE survey. This survey was chosen because it has been used to examine substance use in college-aged persons over a period of 30 years. It was also recommended by one of the authors of, “Raising the minimum drinking age: Some unintended consequences of good intentions,” that lead to the development of the current research (Lanza-Kaduce and Richards, 1989). The CORE survey on drug and alcohol

29

use has been utilized since the 1980s, and was developed by the US Department of Education, in affiliation with several universities. Originally, the CORE survey asked double barreled questions in reference to drug and alcohol use, combining them in each measure, such as “Does your campus have alcohol or drug policies?”. These questions were further developed by separating measures of alcohol, marijuana and other drug use into three separate categories. Questions measuring black market deviance, and black market peer associations were included to further the prior line of research, and there were also questions added to correctly capture the other social learning variables. Skip logic was included in the online survey, to ensure that the students did not spend any unneeded time filling out the survey. If the students chose an answer that made later questions irrelevant, those questions were automatically skipped for them. Thus far, 408 students have responded and are included in this analysis. Data collection is still ongoing, and a maximum of 600 students will be sampled for this phase, however only 408 were included in this analysis. For this thesis, all participants were eighteen and older, and were currently enrolled in classes at the University of Florida. Both male and female students were sampled, and all races were included. Two groups of students were identified through survey questions: students of legal drinking age and students under the legal drinking age. Within these two groups, sub-groups were identified: those who have used persons they are not closely associated with, or strangers, of legal drinking age to procure alcoholic beverages (Alcohol Black Market Sources), and those who have not. The students of legal drinking age were asked to recall the time when they were underage, and were asked if they had ever used alcohol during that time, and if so they were questioned in the same manner as the underage group.

30

Materials and Procedures A survey was administered to measure the variables of interest (see Appendix D). No identifying information was requested from respondents, and it was noted on the survey that all information provided was completely anonymous. Each question required an answer so that students could not skip around in the survey, so every question also had an option listed as “I do not wish to answer.” Responses to survey questions were utilized to determine if there was any relationship between using black market associations to obtain alcohol and using other black market associations. Survey data were also utilized to determine if there was a relationship between black market associations in general and deviance, substance use (not alcohol or marijuana) and specifically black market deviance. The survey data was used to specify measures of the social learning variables, and the outcome behaviors of interest (incidents of underage drinking, incidents of using alcohol black market ties, incidents of using other black market means to commit deviance, etc.). The broader variables were operationally specified as including the various sociodemographic measures of race (white, black, Asian, American Indian, Hispanic, other); education (freshman, sophomore, junior, senior); student status (part time, half time, full time); living status (on campus, off campus); living arrangements (house/ apartment, residence hall, campus housing, fraternity or sorority, other); living situation (live with parents, other students, friends, spouse, children, alone, other); religious affiliation (no religion, cult religion, Jewish, Catholic, Mainline Protestant, Evangelical Protestant); school performance (approximate gpa) and age. These variables were used as control variables.

31

For the current study, 61% of the sample was under 21, 32% were 21-22, and the remaining 7% were 23 years of age or older. Males represented 34.8% of the sample, while females covered 65.2% of the sample. The racial composition of participants was as follows: 64.7% were Caucasian, 14.8% were African American, 14.8% were Hispanic, 3% were Asian/ Pacific Islander and the remaining 2.7% chose other; 97.2% of the sample was single, 97% was also considered full time status as a student, and 25.5% lived on campus, while 74.1% lived off campus. Table 4-1. Participant Characteristics Characteristics Frequency Gender Male 34.8% Female Age Under 21 61% 21-22 32% 23+ 7% Race African 14.8% American Asian/Pacific 3% Islander Caucasian 64.7% Hispanic/Latino 14.8% Other 2.7% Student Status Full time 97% Part time 3% Living Arrangement On Campus 25.5% Off Campus 74.1% In comparison to the demographic makeup of the student body at the University of Florida, females and minorities may have been over represented in this sample. The University of Florida had a student population of 50,912 in the fall of 2006. Females compromised 53% and males 47% of the student body. The minorities are as follows:

32

7.9% are African American, around 11.2% are Hispanic, and about 7% are Asian American or Pacific Islanders (University of Florida-Demographics, nd.). There were two types of substance use questions in this survey, specific and general. For the specific substance use questions the survey clearly indicated one of the following: Alcohol, Marijuana, or Other Drugs and usually asked the same question separately for all three substances. In the past, CORE combined alcohol and drugs into the same question, and did not break out marijuana at all, thus it could not be determined which, if any, of the substances had different usage rates. For the general substance use questions, the survey indicated a list of substances: Tobacco, alcohol, marijuana, cocaine, amphetamines, sedatives, hallucinogens, opiates, inhalants, designer drugs, steroids, or other illegal drugs, this list will be hence forth referred to as the general substance use list. General deviance was also used throughout the survey. It included 14 measures on behavior violating laws/ social norms: damaged property, getting into a physical altercation, driving under the influence, missing class, stealing things worth less than $50.00, stealing things worth more than $50.00, buying stolen goods worth less than $50.00, buying stolen goods worth more than $50.00, pawning or selling stolen goods worth less than $50.00, pawning or selling goods worth more than $50.00, getting into a fight serious enough to cause injury, cheating on an exam or assignment, selling illegal drugs, making unwelcomed sexual advances toward someone, and using a substance to obtain sex. Responses were set up differently depending on the question. This list will hence forth be termed the general deviance list. Measures of Control Variables All included variables are summarized in Table 4-2. Control variables included in all analyses consisted of Gender and Race. Gender was coded as female (0), and male 33

(1). Race included five classifications: African American (1), Asian/ Pacific Islander (2), Caucasian (3), Hispanic (4), and other (5). For ease of analysis, and due to nonsignificance, Race was set up under a dummy variable, and coded as either Caucasian (1) or non-Caucasion (0) (see Table 1. Descriptive Statistics in Table 4-2, page 46). Other control variables were available, but were not significant in any model, so were not utilized for model assessment. Measure of Alcohol Black Market Sources The Alcohol Black Market Sources measure was created by scaling questions designed to determine how the participant was obtaining the alcohol. Four questions were included in the scale: “I have asked someone I don’t know to buy it [alcohol] for me, I have paid someone I don’t know to buy it for me, I asked a student I am not really associated with to buy it for me, and I have paid a student I am not really associated with to buy it for me.” Answer choices included: no (0), yes (1) and I do not wish to answer. Those who chose I do not wish to answer were coded as missing. Because this was a scale created from dichotomized items, the Kuder-Richardson Formula 20 (KR-20) test of internal consistency reliability was utilized rather than a Chronbach’s alpha. The KR20 is designed for dichotomous item scales and it is interpreted the same as a Chronbach’s alpha. The Alcohol Black Market Sources scale was assessed and was found reliable (KR-20= 0.762). Measures of Dependent Variables The measure of Alcohol Black Market Sources given above was utilized as a dependent variable in only one model, a model that used social learning variables specific to drinking and underage drinking behavior (see Table 2, Model 1). As stated above, the measure of black market sources or associations are simply reports of contact or 34

interaction by the person with others for the specific purpose of using them to help obtain, or as a source for, the substance desired. Also, all of the black market measures were different from the differential association measures. Black Market Sources were measured only using strangers, person’s not known to the participant and students not known to the participant. Alcohol Black Market Sources were not used as measures of differential association in this study. Marijuana Use To measure Underage Drinking and Marijuana Use, four questions were asked: “In the following series of questions you will be asked in what ways you obtained alcohol while underage. If you have never obtained alcohol while underage (if you are underage now, or drank back when you were underage), please indicate so below.” Answer options included: I never drank while underage (0); At some point in time, I drank while under the legal drinking age (1); and I do not wish to answer (missing). Similar questions were asked for marijuana: “In the following series of questions you will be asked in what ways you have obtained [marijuana, other drugs]. If you have never used [marijuana, other drugs], please indicate so below.” Answers were dichotomized as: I have never used [marijuana, other drugs] (0), At some point in time, I have used [marijuana, other drugs] (1), and I do not wish to answer (missing). Since these were single item measures, no scale was created. Marijuana Black Market Sources For the question, do Alcohol Black Market Sources lead to involvement with other black market sources, a scale for Marijuana Black Market Sources. If the participant answered that they had, at some point in time used marijuana, skip logic automatically progressed them to this question: “…have you used the following ways to obtain 35

[marijuana/other drugs]?”. There were four items measuring black market sources for both marijuana and other drugs: I buy it from someone a friend introduced me to, I buy it from someone I don’t know, I buy it from a student I am not really associated with, and I have attempted to buy it without a contact. The answer choices were dichotomized for these scales with options including no (0), yes (1) and I do not wish to answer was treated again as missing. To create the scale, items were summed and then coded into a dummy variable including no (0 remained 0) and yes (all other values were recoded to 1, indicating that they have utilized that method of obtaining the substance). This scale was assessed with the KR-20 (Marijuana Black Market Sources yielded a KR-20 of 0.818). General Deviance The scale for General Deviance was constructed using the general deviance list (noted above). Answer items were set up on a six item ordinal scale, ranging from never (0) to ten or more times (5), and I do not wish to answer (missing). Scale reliability was assessed and yielded a Cronbach’s alpha of 0.729. There are some limitations of this study in using General Deviance as an outcome variable. Specifically, the concepts of differential reinforcement and imitation were not measured for General Deviance, but only for use of alcohol, marijuana. There was one set of deviance specific questions that measured attitude and that was Risk Deviant Behavior (see below, social learning measures, Definitions/ Attitudes). To be more accurate measures of social learning variables as independent variables in General Deviance, it would be better to have these explanatory variables measured specifically with regard to general deviant behavior. Specific Black Market Deviance If a participant answered that they had, at some point in their lives, used [alcohol/ marijuana] they were asked in what ways they obtained it. To measure the relationship 36

between the method of obtaining a substance and deviance associated with it, a series of questions was asked. Of interest is the black market deviance (Specific Black Market Deviance) that is associated directly with Black Market Sources, measured in the following: “Did any of the following occur when you had a stranger get you [alcohol/ marijuana]?” and the 14 general deviance measures were listed. Next, “Did any of the following occur when you had a student you don’t really know get you [alcohol/ marijuana]?” and the 14 general deviance measures were listed. In addition, for marijuana and other drugs, it was also asked “Did any of the following occur when you tried to get [marijuana] without a contact?” and again the 14 general deviance measures were listed. Answers for each of these questions were set up on a binary yes/no (1/0). Within the general deviance measures, there are 5 questions that indicate black market deviance. They are: “you bought stolen goods worth less than $50.00, you bought stolen goods worth more than $50.00, you pawned or old stolen goods worth less than $50.00, you pawned or sold stolen goods worth more than $50.00, and you sold illegal drugs.” The responses for the questions (yes/no) were scaled to indicate black market deviance specifically associated with utilizing black market sources (Specific Black Market Deviance). The reliability of Specific Black Market Deviance was tested, and yielded a KR-20 of 0.988. There was also a measure of deviance in general (General Black Market Deviance), that included deviant behavior that was specific to the times the participants used black market sources to obtain alcohol, but was not considered black market deviance (such as dealing in stolen goods). For this scale, the deviance questions (listed above) specific to Black Market Sources were summed and recoded into a dummy variable including no (0 remained 0) and yes (all other values were recoded to 1,

37

indicating that they have committed deviance while obtaining alcohol via black market sources). Deviance specific to Alcohol Black Market Sources yielded a KR-20 of 0.980 (Specific Black Market Deviance). Due to the low response rate, validity of the models analyzing Specific Black Market Deviance was questioned, and thus these models were dropped from the analysis. It will be interesting to see if the response rate for these questions increases with larger and broader sample sizes in the next study. The models with these dependent variables were analyzed with Alcohol Black Market Sources as the only independent variable, and then the next model shown in the tables added the measures of the social learning variables (see below). Recall, that it was hypothesized that when entered into the equation, the social learning variables would explain most of the variance in Marijuana use, Marijuana Black Market Sources and General Deviance. Again, since the use of marijuana most likely occurred after the initial use of alcohol, the expectation was that use of Alcohol Black Market Sources preceded and created a social context which led to a higher probability of using Marijuana Black Market Sources, Marijuana Use and Deviance. However, since all of these were hypothesized to result from the same underlying social learning process it was expected that when measures of the social learning variables of differential reinforcement, definitions, differential association and imitation were entered into the model, the net effects of the use of black market sources for alcohol (Alcohol Black Market Sources) on these other forms of deviance would be lower than their main effects and be statistically non-significant.

38

Measures of Social Learning Variables Differential Reinforcement Differential reinforcement was measured in several different ways. As previously stated, the balance of reinforcement and punishment, both fall under the category of differential reinforcement. So, differential reinforcement by definition includes measures of both reinforcing and punishing stimuli, as it is the balance of reinforcement and punishment that yields the maintenance or desistance of a particular behavior. Consequences were measured for alcohol use and marijuana use by asking, “From what you have experienced, have the consequences of your substance use been overall:”. Answers were as follows: I don’t use this substance (0), more bad than good (1), about the same for good/bad (2), more good than bad (3), and I do not wish to answer (missing). Because these were single item measures, no scale was created (Alcohol Consequence, Marijuana Consequence). After thesis defense, the Consequence items were dropped from all models, and are not reported in any of the findings, per committee members’ request. Potential punishment (Alcohol Punish and Marijuana Punish) was measured in a question on experiences with alcohol and marijuana (reported separately) using a seven item inquiry, “Please indicate how often, if ever, the use of [alcohol/ marijuana] has lead to each of the following experiences:” The items included: been criticized by someone I know, been hurt or injured, been the victim of unwanted sexual intercourse, done something I later regretted, endured threats of physical violence or actual physical violence, missed a class, performed poorly on a test or important project. Answers were offered on a six item ordinal scale ranging from never (0) to 10 or more times (6). “I do not wish to answer” was coded as missing. These questions were asked both of students 39

who had used each substance, and students who had not used them. Participants indicating that they had not used [alcohol/ marijuana] were asked, “You indicated you have never used [alcohol/ marijuana]. In this question, please imagine how often, if ever, you think the use of [alcohol/ marijuana] would lead to the following experiences.” The experiences and answers were the same. To scale these items, first the questions for the users and non-users were combined, to yield one measure for punishment for each substance. Then all seven items were summed and divided by seven to create the punishment scales for each substance (Alcohol Punish yielded a Cronbach’s alpha of 0.880, Marijuana Punish yielded a Cronbach’s alpha of 0.923, and Other Drug Punish yielded a Cronbach’s alpha of 0.972). Differential reinforcement was also measured by items requesting information on positively viewed effects of alcohol, marijuana and other drugs, individually (Alcohol Reward, Marijuana Reward). This question also included seven items, but for possible positive outcomes, and was listed as: “Do you think [alcohol/ marijuana] have any of the following effects?” The items included were: It is an ice breaker, It enhances social activity, It makes it easier to deal with stress, It facilitates a connection with peers, It allows people to have more fun, It makes me sexier, It facilitates sexual opportunities” with answers set up on a binary response (yes/no), and I do not wish to answers were coded as missing. A scale was then created for Alcohol Reward (KR-20=0.733) and Marijuana Reward (KR-20 = 0.787) by summing the responses and dividing by seven. Definitions/ Attitudes The social learning concept of definitions favorable and unfavorable to substance use was measured with several types of questions throughout the survey. The first inquires about attitudes on using alcohol (Attitude Alcohol) and marijuana (Attitude 40

Marijuana)(each asked separately): “Which statement best represents your own attitude about the following substance use: [Alcohol/ Marijuana]?” Answers were listed on a 5 item Likert scale ranging from strongly disapprove (1) to strongly approve (5). These items were not scaled and used respectively as measures of the variables shown in the tables as Attitude Alcohol and Attitude Marijuana. Participants were also asked about their preferences of availability of alcohol, marijuana and other drugs on campus: “There are differing views with regards to the availability of alcohol and drugs at parties on or around campus. Some students believe that the availability of drugs and alcohol is a bad thing, decreasing their enjoyment and leading to negative situations. Other students think that having drugs and alcohol at these parties increases enjoyment and is a positive thing. Which view would you say you are most like to side with? I would rather have [alcohol, marijuana, other illicit drugs]:” Answer options included not available (0), Available (1), and I do not wish to answer (treated as missing). This question was asked individually for Available Alcohol and Available Marijuana. Because availability of alcohol, marijuana and other drugs were single item measures, no scales were needed. This variable [Available Alcohol/ Available Marijuana] was subsequently dropped from the models, as it only reached statistical significance in a few of the models and was deemed as unnecessary for overall explanatory power. The last Definitions/Attitudes measure included in the analyses was perceived risk (Risk Heavy Drinking, Risk Marijuana, Risk Deviant Behavior). The questions asked, “By partaking in the following activities, how much do you think people risk harming themselves? (physically, mentally, spiritually, emotionally)”. The question was asked for each item on the general substance use list (see above for an explanation of the general

41

substance use list), and answers were set up on a Likert scale, ranging from no risk (0) to great risk (4). Scales were created for Risk Heavy Drinking (Cronbach’s alpha = 0.770) by summing the three alcohol items and dividing by three, and Risk Marijuana (Cronbach’s alpha = .892) by summing the three marijuana items and dividing by three. Questions were also asked about risky/ deviant behaviors (Risk Deviant Behavior) such as: “How much to do you think people harm themselves by: consuming alcohol prior to being sexually active, using marijuana prior to being sexually active, using other drugs prior to being sexually active, regularly engaging in unprotected sexual activity with multiple partners”. Answers were set up on the same ordinal scale as the other risk questions. The four items were summed and divided by four to create the Risk Deviant Behavior variable (Cronbach’s alpha = 0.828). Differential Association The concept of differential association was captured via several different questions in the survey instrument. Family substance use was measured for both alcohol and marijuana separately. The same question was asked for alcohol (Family Alcohol) and marijuana (Family Marijuana): “Which members of your family, if any, have ever had problems with [alcohol/ marijuana]?” Answers included: none, mother, father, stepmother, stepfather and sibling, and were originally coded from 0-5. Dummy variables were created for each relative so that a 0 indicated no problem and a 1 indicated that the relative had a problem with the substance. Items were then summed on a five item scale, and again dichotomized to indicate that either no family members has had an issue with the substance (0), or that at least one of their family members had experienced a problem with the substance (1). The reliability of these two scales was assessed and

42

Family Alcohol yielded a KR-20 of 0.46, Family Marijuana yielded a KR-20 of 0.50. These were the least reliable scales constructed. Deviance of peers was measured in several ways:, friend general deviance (Deviant Peers), Friend Substance Use, Friend Alcohol Consequence, Friend Marijuana Consequence, and Friend Buy Alcohol Under the LDA. Friends’ general deviance (Deviant Peers) was measured using the general deviance list (see above for question details in general deviance list), and the question was posed, “Please indicate how many of your friends have ever experienced the following:”. Answers included: none of my friends (0), a couple of my friends (1), a few of my friends (2) half of my friends (3), a majority of my friends (4), all of my friends (5) and I do not wish to answer (treated as missing). Items were scaled to create the deviant peers measure (Chronbach’s alpha = 0.912). These were the only three differential association measures utilized in this study. For Friend Substance Use the question read: “Have your friends ever told you they have: smoked marijuana once or twice, smoked marijuana occasionally, smoked marijuana regularly, tried cocaine once or twice, done cocaine regularly, tried LSD once or twice, done LSD regularly, tried amphetamines once or twice, done amphetamines regularly, tried club drugs once or twice done club drugs regularly, tried antidepressants once or twice, done antidepressants regularly, tried prescription drugs once or twice, done prescription drugs regularly, Had one or two alcoholic drinks nearly every day, had four or five alcoholic drinks nearly every day, had five more drinks in one setting, purchased alcohol while underage, taken steroids for body building or improved athletic performance. The answers were coded on a binary response (yes/no) and “I do not wish to answer” was treated as a missing variable. For this scale, “purchased alcohol while

43

underage” was removed and treated as a separate item indicating friend deviance (See Table 4-2 variable “Friend Bought Alcohol Under LDA”). The response was dichotomous. Also, all questions for each substance were first summed, and then recoded into a dummy variable, before scaling. So there are three friend substance use variables, Friend Problem Drinking, Friend Marijuana Use, and Friend Substance Use. For example, there are three questions on Marijuana (“Have your friends ever told you they have: smoked marijuana once or twice, smoked marijuana occasionally, smoked marijuana regularly?”). All three marijuana questions were summed, and then recoded into no (0 remained 0) and yes (all other values were recoded to 1, indicating that their friends have told them they have used marijuana, but not indicating to what degree). This was done so that duplicate measures of each substance were not included in the scale, but yet each item was still accounted for. Once all substance questions were collapsed in this manner, a scale was created for friends’ substance use (KR-20 = 0.825). Friend consequences were measured for alcohol use and marijuana use (Friend Alcohol Consequence, Friend Marijuana Consequence) by asking, “From what you have observed, or know about from your friend’s, have the consequences of their [alcohol/ marijuana] use been overall:”. Answers were as follows: My friends don’t use this substance (0), more bad than good (1), about the same for good/bad (2), more good than bad (3), and I do not wish to answer (missing). Because these were single item measures, no scale was created (Friend Alcohol Consequence, Friend Marijuana Consequence). These consequence measures were also dropped from each model at the behest of committee members, post thesis defense.

44

Imitation/ Modeling The final social learning concept, Imitation, is measured with one series of questions, and a single item question. “Have you ever directly observed your friends: smoking marijuana once or twice, smoking marijuana occasionally, smoking marijuana regularly, trying cocaine once or twice, doing cocaine regularly, trying LSD once or twice, doing LSD regularly, trying amphetamines once or twice, doing amphetamines regularly, trying club drugs once or twice doing club drugs regularly, trying antidepressants once or twice, doing antidepressants regularly, trying prescription drugs once or twice, doing prescription drugs regularly Having one or two alcoholic drinks nearly every day, having four or five alcoholic drinks nearly every day, having five more drinks in one setting, purchasing alcohol while underage, taking steroids for body building or improved athletic performance.” The answers are coded on a binary yes/no scale with the option “I do not wish to answer” coded as missing. “Purchasing alcohol while underage” was not included in the scaled items, and was treated as a separate measure of imitation, as a single item question. The three alcohol questions was also not included in this scale, but in a separate scale. To capture imitation of substance use, measures for each substance were first summed, and then coded into a dummy variable. For example, there are three questions on marijuana (“Have you ever observed your friends: smoking marijuana once or twice, smoking marijuana occasionally, and smoking marijuana regularly?”). All three marijuana questions were summed, and then recoded into no (0 remained 0) and yes (all other values were recoded to 1, indicating that they have observed their friends using marijuana, but not indicating to what degree). This was done so that duplicate measures of each substance were not included in the scale, but yet

45

each item was still accounted for. Once all substance questions were collapsed in this manner, a scale was created for Observed Friend Substance Use (KR-20 = 0.691). For Model 1 only (see Table 5-1, p. 54) a measure on observation of friend problem drinking (Observed Friend Problem Drinking) was also included. This measure was a three item scale created from questions that asked specifically about observation of friend problem drinking. These questions asked if the person had observed their friends: “Having one or two alcoholic drinks nearly every day, having four or five alcoholic drinks nearly every day, having five more drinks in one setting, purchasing alcohol while underage,” Answers were dichotomized on a yes/no response. (KR-20 = 0.672). This variable was not used in later models to eliminate tautology issues with the Alcohol Black Market Sources variable, as it was subsequently used in the remaining models as an independent variable. It is recognized that the empirical measure of imitation of peers has presented some problems in that its measures and effects are difficult to separate from those of differential peer association and that may be implicated in the relative strong effects of imitation in these findings compared to the effects of imitation in some previous research (Akers, 1998). Asking about directly observing peers engaged in various acts is, of course, a clear indicator and reflection of the concept of imitation as observational learning. But since one may not directly observe the behavior of someone with whom one is in face-to- face contact without in effect associating with that person, it also may be seen as partly a reflection of the concept of differential peer association.

46

Table 4-2. Descriptive Statistics Variables Observations Dependent variables *Alcohol Black Market Sources 335 Marijuana Black Market Sources 154 General Deviance 387 Marijuana Use 373 Control Variables Gender 408 Race 405 Differential Reinforcement Alcohol Reward 386 Marijuana Reward 382 Alcohol Punish 405 Marijuana Punish 399 Friend React Heavy Drinking 404 Friend React Marijuana 403 Friend React Substance Use 402 Definitions/ Attitudes Risk Heavy Drinking 406 Risk Marijuana 406 Risk Deviance 406 Attitude Alcohol 407 Attitude Marijuana 407 Differential Association

Standard Mean Deviation Minimum Maximum 0.50 0.25 0.32 0.42

0.99 0.43 0.34 0.49

0 0 0 0

1 1 7 1

0.35 3.73

0.47 0.44

0 0

1 1

4.58 2.55 0.87 0.89 2.15 2.34 1.48

1.78 2.10 0.93 1.11 0.90 0.93 0.72

0 0 0 0 1 1 1

7 7 5 5 5 5 5

2.15 1.41 2.18 3.51 2.43

0.71 0.88 0.73 0.91 1.10

0 0 0 1 1

3 3 3 5 5

Family Alcohol 408 0.34 0.67 0 1 Family Marijuana 408 0.21 0.56 0 1 Deviant Peers 377 0.38 0.28 0 7 Friend Substance Use 393 3.52 2.31 0 8 Friend Buy Alcohol Under LDA 398 0.78 0.47 0 1 Imitation/ Modeling Imitation 396 2.19 1.78 0 9 Imitation Friend Buy Alcohol Under LDA 397 0.55 0.50 0 1 Observed Friend Problem Drinking 396 0.71 0.46 0 1 * Alcohol Black Market Sources is used as a dependent variable in Model 1, Table 5-1 only, in all subsequent models it is an independent variable.

Data Analysis The analyses were completed using SPSS version 17 for graduate students and Stata version 10 for graduate students. Descriptive statistics for all the variables that were used, as well as frequencies, were run to determine case counts and variance (See

47

Table 4-2). The analysis and reporting of the data was quantitative by design. The scales and measures explained above were utilized in a series of either logistic regression or ordinary least squares regression depending on whether or not the dependent variable was dichotomous. Models that included dichotomized outcome variables were assessed via logistic regression as not to violate any of the assumptions for linear regression. The logit model is a method of estimation for equations with dummy variables, or binary response variables. Utilizing the logit model, “avoids the unboundedness problem of the liner probability model…the dependent variable… can be thought of as the log of the odds that the choice in question will be made” (Studenmund, 2001: 442). Because logit models report natural log odds as the coefficient, the percent change was calculated so that the coefficients were easily interpreted. The odds is the ratio of the probability that something is true divided by the probability that it is not true. Before any models were conducted, data were checked for outliers. There were no cases that appeared to be problematic. Listwise deletion was used for missing data. Of the 408 participants, only 335 had (at some point in their lives) drunk while under-age. Of those 335, 302 participants answered all of the questions that were included in Model 1 (Page 54). Ordinary least squares regression was completed for the first model (Model 1, see Table 5-1 page 54). In this model being involved with Alcohol Black Market Sources was predicted with a set of social learning variables that were specific to drinking and underage drinking behavior. The social learning variables assessed in this model included: attitudes favorable or unfavorable towards alcohol (Attitude Alcohol), whether or not their friends told them they had purchased alcohol while under the LDA (Friend Bought Alcohol Under LDA), whether they had seen their

48

friends purchase alcohol under the LDA (Imitation Friend Buy Alcohol Under LDA), whether their friends told them they had problem drinking behaviors (Friend Problem Drinking - i.e. binge drinking), whether they had seen their friends partake in problem drinking behaviors (Imitation Friend Problem Drinking), whether they had seen their friends drink (Imitation), how they thought their friends would react to their own problem drinking behaviors (Friend React Heavy Drinking), whether or not their drinking behavior had been rewarded (Alcohol Reward), whether or not their drinking behavior had been punished (Alcohol Punish, in the behavioral sense of punishment), whether they felt the consequences of alcohol were good or bad (Alcohol Consequence), family alcohol use (Family Alcohol), consequences of friend alcohol use (Friend Alcohol Consequence) and finally a measure of Deviant Peers was included. The model was tested for heteroskedasticity using the Shapiro-Wilk test for normal data, it was determined that the residuals were normally distributed, so there were no issues with heteroskedasticity. Collinearity diagnostics were run on all the variables included in this model, and there were no issues with collinearity or tolerance. Leverage was tested for by calculating the leverage cut off value (2x13/302= .086). There were few cases observed over the leverage cut off, and so these cases were not removed. The DFbeta was created by dividing 2 by the square root of 302, yielding a cut off value of .115. There were several cases noted in each variable that were over the cut off, however the number of cases was low and this author did not think the outcome would be substantially affected by the removal of so few cases. After thesis defense, committee members felt it wise to delete the variables Alcohol Consequence and Friend Alcohol Consequence, and so these variables do not appear in the new analyses.

49

For the next two models, a series of logistic regression was conducted (see Table 52 on page 56). Of the 335 participants who had, at some point in their lives, used alcohol while under the legal drinking age, 151 had used Marijuana Black Market Sources. In Model 2 there are 151 observations reported. In Model 3, 135 of those 151 participants answered all the questions and were included in that analysis. Both of these models were also run using only the 135 participants included in Model 3 of the original analysis; however; results did not differ, so the original analysis is reported here. The first model in this series (Model 2) utilized Alcohol Black Market Sources, along with control variables, to predict contact with Marijuana Black Market Sources. The second model in this series (Model 3) included the same outcome variable (Marijuana Black Market Sources), but contained measures of the social learning variables (different from those used in model 1), along with the control variables and Alcohol Black Market Sources as independent variables. The social learning variables assessed in this model included: Family Marijuana, Risk Marijuana, Imitation, Attitude Marijuana, and Friend Marijuana Consequence. Collinearity diagnostics indicated no issues of tolerance or collinearity. Percent change in odds was also calculated for each variable in both models. The same series of models was attempted for the dependent variable Other Drug Black Market Sources, however the response rate was not high enough to calculate the models. Under the Definition/Attitude portion of the model, there is not a measure of the perceived risk of alcohol use, thus Risk Alcohol was not included in this model. Note, in these models Friend Buy Alcohol Under LDA and Imitation Buy Alcohol Under the LDA were dropped as they were not relevant to these analyses. (Also, Marijuana Consequence, and

50

Friend Marijuana Consequence were removed as suggested by committee members, and are not reported in these analyses. The next series of models (see Table 5-3, page 58) focused on the use of marijuana (Marijuana Use) as the outcome variable of interest (Models 4-5). Model 4 utilized Alcohol Black Market Sources, controlling for Race and Gender, to predict Marijuana Use (see Table 5-3, page 59). Of the 335 participants who had drank while underage, 316 had used marijuana (Model 4). Of those 316 people, 281 participants answered all the questions required for Model 5. (These two models were run again, using only the 281 participants from Model 5. There were no differences in the results in significance, and thus original analyses are reported here). Model 5 included the same variables, but also had the social learning variables specific to Marijuana Use. Collinearity diagnostics indicated no issues of tolerance or collinearity. Percent change in odds was also calculated for each of the models. This author originally ran two models predicting Other Drug Use utilizing the same variables listed above, but due to a low response rate, the validity of the significance was questionable, and thus those two models were dropped. Note, in these models Friend Buy Alcohol Under LDA and Imitation Buy Alcohol Under the LDA were dropped as they were not relevant to these analyses. Next, two linear regression models were conducted (see Table 5-4, page 60). The first, Model 6, was to assess the relationship between Alcohol Black Market Sources, control variables, and General Deviance. The second, Model 7, included the same variables, but also included social learning measures. Of the 335 participants who had used alcohol while underage, 322 had participated in some sort of general deviance

51

(Model 6), and of those 322 participants, 295 participants answered all the questions to be included in Model 7. (Post thesis defense, again based on suggestions of committee members, these two models were run a second time, utilizing only the 295 participants that answered all the questions. There was virtually no difference in the results, and no difference in the statistical significance, so the original models were reported here.) The variable, Imitation Friend Buy Alcohol Under the LDA, was included in this model because it was observed deviance, and thus relevant to the analysis. The learning variables utilized in these models were somewhat weaker in that only two series of questions specifically targeted deviance (Deviant Peers and Risk Deviant Behavior) throughout the survey. There were no other questions that assessed attitudes/definitions towards deviance, and there were no direct measures of imitation of deviance. Substance use measures were utilized however, and were still valid predictors. The model was tested for heteroskedasticity using the Shapiro-Wilk test for normal data, it was determined that the residuals were not normally distributed. To account for this lack of normal distribution, robust standard errors were calculated. In comparing the p-values between the two, it was determined that there was a difference in the standard errors, and thus the robust standard errors are reported. Collinearity diagnostics were run on all the variables included in this model, and there were no issues with collinearity, there were also no issues with tolerance. Leverage was tested for by calculating the leverage cut off value (2x9/299= .06). There were only 14 cases observed over the leverage cut off, and so these cases were not removed. The DFbeta was created by dividing 2 by the square root of 299, yielding a cut off value of .116. There were several cases noted in each

52

variable that were over the cut off, however the number of cases was low and this author did not think the outcome would be substantially affected by the removal of so few cases. This researcher had planned to conduct analyses testing the relationship between Alcohol Black Market Sources and both General Black Market Deviance, and black market deviance specific to utilizing black market sources to obtain illicit substances (Specific Black Market Deviance). The response rate for these items, however, was too low to accurately predict any type of relationship, and thus these analyses were dropped.

53

CHAPTER 5 RESULTS Alcohol Black Market Sources Table 5-1. Model 1 Variables

Model 1 b

Control Variables Gender 0.02 Race - 0.07 Differential Reinforcement Alcohol Reward 0.07 Alcohol Punish 0.18 Friend React Heavy Drinking 0.11 Definitions/ Attitudes Attitude Alcohol 0.13 Risk Heavy Drinking 0 .05 Differential Association Family Alcohol -0.09 Deviant Peers -0.12 Friend Buy Alcohol Under LDA -0.15 Imitation/ Modeling Observed Friend Problem Drinking -0.28 Imitation Buy Alcohol Under LDA 0.05 Imitation 0.18 2 F = 6.49 n=302 R = .23

SE

t

0.12 0.04

0.14 -1.58

0.04 0.08 0.07

1.86* 2.29** 1.56*

0.07 0.09

1.75* 0.55

0.07 0.21 0.17

-1.17 -0.57 -0.86

0.16 -1.78* 0.14 0.40 0.04 4.51*** *P ≤ .10, ** P ≤ .05, *** P ≤ .01

Model 1: Ordinary Least Squares Regression Predicting involvement with Alcohol Black Market Sources with Social Learning Variables

The first model was assessed via ordinary least squares regression. The act of obtaining alcohol through the black market (Alcohol Black Market Sources) is considered deviant and generally perceived to precede procurement and use of other illicit substances. Alcohol Black Market involvement (Sources) was first predicted with social learning variables, as measured with regard to specific drinking behaviors. Model 1 achieved statistical significance (p-value